我上一篇刊登于《东方日报》的文章〈美国对中国人工智能不安可引向核战〉，指出了人工智能竞争是“末日钟”倒数中的一项关键发展。
这涉及一个明确进入主流政策论述的重大变化：在美中争夺通用人工智能（AGI）的竞赛中，动能行动——例如对半导体晶圆厂和人工智能数据中心发动传统导弹袭击或进行实体破坏——已被纳入政策讨论之中，这正是新美国安全中心（CNAS）的印太安全项目副主任雅各布・斯托克斯（Jacob Stokes）所提出的主张。
虽然西方主流媒体的评论经常探讨人工智能如何可能导致灾难性或“末日”般的后果，但他们却系统地避开，或重新包装“国家授权对敌方科技基础设施发动动能打击”这一概念。
因此，有必要检视《纽约时报》、《华尔街日报》、《金融时报》、英国广播公司（BBC）和美国有线电视新闻网（CNN）等西方主流媒体，在这个议题上所呈现的主要论述，并将其与本人在《东方日报》文章所提出的分析框架进行对照。
西方媒体如何框定人工智能灾难风险
当西方评论员、政策分析人士和媒体讨论人工智能走向灾难的可能性时，其论述通常可分为三大类：
生存风险与失控对齐：这类论述集中于超级智能系统脱离人类控制、采取与人类价值不一致的行动，或者在无意中将有害目标最佳化的技术威胁。
自主军事升级与自动化指挥：这类论述关注机器可能造成的误判——也就是在危机期间，若将人工智能体整合进战略指挥与控制流程、预警系统或自主武器之中，可能导致快速且非预期的局势升级。
地缘政治“算力遏制”与供应链冲击：这类论述则透过民用经济工具来描述科技战，包括出口管制（美国工业与安全局的限制）、晶片封锁、对外投资审查，以及针对国家支持骇客的防御性网络安全措施。这些措施都是针对中国。
为何西方媒体避谈“动能打击”选项
主流西方评论没有触及一个关键的核心问题——对敌方人工智能中心发动先发制人的军事打击——其背后存在战略、修辞、制度与意识形态上的界线：
跨越国际法门槛：
主张或将导弹攻击核武国家（中国）境内的商业数据中心或军民两用设施正常化，将构成直接的战争行为。虽然像新美国安全中心（CNAS）这类智库政策提案，将动能选项与网络及外交手段并列为理论上的最终情境，但主流西方媒体认为，对外国主权基础设施发动实体打击过于激进或危险，因此不会公开倡议。这样做将破坏围绕国家主权建立的国际法律规范。
防御与预防之间的不对称框架：
西方媒体在报导科技基础设施遭受实体攻击时，几乎完全聚焦于防御韧性，而不是进攻性的预防措施。例如，在2026年3月中东亚马逊云端基础设施遭无人机袭击后，相关评论将数据中心描述为容易遭受攻击的目标，需要实体防御、反无人机系统以及强化主权云端安全。西方媒体很容易分析科技基础设施如何成为敌方侵略的目标，却抗拒将西方国家描述为可能对外国科技中心发动先发制人攻击的一方。
“经济竞争”的修辞盾牌：
只要把公共讨论维持在贸易制裁、出口管制（英伟达、超微AMD、ASML的光刻技术）以及透过《晶片法案》推动供应链回流等范畴，西方评论员就能维持“经济与监管遏制”的叙事。承认“算力剥夺”可能需要透过实体摧毁来实现，将暴露经济制裁的局限，也意味著非军事遏制策略最终可能不足以达成目标。这将为军事手段或其他形式的肮脏非军事攻击打开大门。
不同分析
显而易见，西方评论将人工智慧技术应用、军事战术升级和产业贸易政策隔离开来，以免引起公众恐慌。
重点放在“智库到信条”的转化路径：
我们不应低估新美国安全中心（CNAS）等具影响力政策中心所发出的论述。这包括其在《超级大国与通用人工智能》、《红线》以及《对抗蒸馏》等报告。与其把“打击数据中心”或“动能选项”视为遥不可及的科幻情境，我们更应该看到，华盛顿主流政策规划正在发生转变，开始把中国的算力中心视为具有正当性的军事目标。
将AGI从商业问题重新界定为核力量平衡问题：
西方媒体的文章往往把人工智能硬件视为具有军民两用性质的商业资产。但如今的现实是，华盛顿正从关乎生存的军事能力角度看待AGI门槛的突破——类似一个国家跨越核武门槛。这种转变从根本上改变了华盛顿对先发制人行动的门槛。
全球南方／非西方的地缘政治现实主义：
中国采取的结构性反制措施——例如“数码丝绸之路”、稀土出口咽喉要道，以及将数据中心向西部省份进行实体加固和分散部署——显示，遏制并非孤立发生，而是会引发双方危险的、自我强化的强化循环。
预判美中人工智能竞争
分析美中人工智能动态的分析人士、政策制定者和战略家，也应将以下因素纳入考量。
选定打击目标的早期预警指标：
值得注意的是，围绕针对算力基础设施采取动能攻击及进攻性网络行动的政策论述，已经进入主流军事兵棋推演。这是一项早期预警：数据中心、海底电缆以及晶圆厂的能源电网，正在从商业基础设施转变为军事规划中的主要目标。
揭示“遏制陷阱”：
积极的出口管制（BIS的关键瓶颈措施）迄今未能冻结中国的技术能力。相反，这些措施迫使中国加速推进全产业链自主化，包括华为昇腾和中芯国际（SMIC），以及采取不对称的替代方案，例如提高算法参数效率和发展具身工业机器人。
降级紧张的诊断指标：
应该制定一套现实的指标，并优先用于判断美中人工智能关系究竟是在改善还是恶化。如果政策制定者继续把商业前沿大型语言模型（LLM）及AGI的发展，与攸关生存的反制力量威胁混为一谈，那么任何关于“人工智能核指挥与控制”（AI nuclear C2）的正式双边协议，或者共同安全准则，都将无法实现。
结论
西方媒体和政策圈目前呈现出一个显著脱节，内部国防政策规划与面向公众的媒体论述之间存在明显落差。
当西方私人企业及媒体主导的评论担心由算法故障引发“意外”人工智能战争时，它们却一再忽略另一种风险——有意识的传统军事先发制人行动。这种军事决策可能源自于恐惧敌人距离取得无法匹敌的AGI突破，可能只剩几个月。
如果AGI被视为一种攸关生存的威胁，足以永久决定全球力量平衡，那么先发制人的反制打击，其战略逻辑将与冷战时期的核武战略相似。然而，这一次却缺乏确保“末日时钟”不会走到12点所必需的安全保障机制。
林德宜《美国会因人工智能竞赛而摧毁中国吗？》原文：Will the US Take Down China Because of AI?
My last Oriental article highlighted a critical development in the countdown of the Doomsday clock arising from AI rivalry. This relates to the explicit mainstreaming of kinetic actions—such as conventional missile strikes or physical sabotage against semiconductor foundries and AI data centers—into the policy discourse regarding the U.S.–China race for Artificial General Intelligence (AGI), as proposed by Jacob Stokes of the Center for a New American Security (CNAS).
While mainstream Western media commentaries frequently examine how AI could lead to catastrophic or "doomsday" outcomes, they systematically have avoided or reframed the concept of state-sanctioned kinetic strikes against adversary tech infrastructure. It is useful to examine the prominent themes running through mainstream U.S. and European outlets (such as The New York Times, The Wall Street Journal, Financial Times, BBC, and CNN) on the subject and contrast them with the analytical framework proposed in the Oriental article.
How Western Media Frames AI Catastrophic Risk
When Western commentators, policy analysts, and media outlets discuss AI's path to catastrophe, their narratives generally fall into three distinct categories:
Existential Risk & Rogue Alignment ("p(doom)"):
Concentrating on the technical threat of superintelligent systems escaping human control, acting alignment-averse, or inadvertently optimizing for harmful goals.
Autonomous Military Escalation & Automated Command:
Focusing on machine miscalculation—how integrating AI agents into strategic command-and-control loops, early-warning systems, or autonomous weapons could trigger rapid, unintended escalation during a crisis.
Geopolitical "Compute Containment" & Supply-Chain Shock:
Framing the technology war through civil-economic instruments: export controls (Bureau of Industry and Security restrictions), chip blockades, outbound investment screens, and defensive cybersecurity against state-backed hackers. These are directed against China.
Why Western Media Avoids the "Kinetic Option" Narrative
The absence of a critical central concern—preemptive military strikes on adversary AI centers— is not mentioned in mainstream Western commentary for strategic, rhetorical, institutional and ideological boundaries:
Overstepping the Threshold of International Law:
Proposing or normalizing missile strikes on commercial or dual-use data centers within a nuclear-armed power (China) represents a direct act of war (casus belli). While think tank policy proposals like CNAS's Superpowers and AGI list kinetic options alongside cyber and diplomatic vectors as theoretical end-state contingencies, mainstream Western outlets treat physical strikes on sovereign foreign infrastructure as too radical or dangerous to advocate publicly, as doing so would destabilize international legal norms around state sovereignty.
Asymmetry in Defensive vs. Preventive Framing:
Western coverage of physical attacks on technology infrastructure focuses almost exclusively on defensive resilience rather than offensive prevention. For instance, commentary following the March 2026 drone strikes on AWS cloud infrastructure in the Middle East framed data centers as vulnerable targets that require physical defense, anti-drone coverage, and sovereign cloud hardening. Western media readily analyzes tech infrastructure as a target of enemy aggression, but resists portraying Western nations as potential preemptive attackers on foreign tech hubs.
The Rhetorical Shield of "Economic Competition":
By keeping the public discussion focused on trade sanctions, export controls (Nvidia, AMD, ASML lithography), and supply-chain reshoring (via the CHIPS Act), Western commentators preserve a narrative of economic and regulatory containment. Admitting that "compute denial" might require physical destruction exposes the limits of economic sanctions and signals that non-military containment strategies may ultimately prove insufficient. This will open the door to military or other forms of dirty non military attacks
How The Edge Analysis Differs
It is clearly evident that Western commentary divides technical AI alignment, military tactical escalation, and industrial trade policy into separate silos so as not to alarm the public.
Key Focus on the "Think Tank to Doctrine" Pipeline:
It is important to not underestimate the discourse originating from influential policy centers like the Centre for a New American Security (CNAS) (referencing their Superpowers and AGI, Red Lines, and Adversarial Distillation reports). Rather than treating "striking data centers" or "kinetic options" as far-fetched sci-fi scenarios, we should be seeing the shift in mainstream Washington policy planning toward treating China's compute hubs as legitimate military targets.
Reframing AGI from Commerce to Nuclear Parity:
Western media pieces often treat AI hardware as a commercial dual-use asset. The reality now is that Washington views AGI threshold attainment through the lens of existential military capability—akin to a nation crossing the nuclear threshold. This shift fundamentally has changed the Washington threshold towards preemptive action.
Global South / Non-Western Geopolitical Realism:
China’s structural countermeasures—such as the Digital Silk Road, rare earth export chokepoints, and the physical hardening/dispersion of data centers into western provinces—shows that containment does not happen in a vacuum, but triggers a dangerous, self-reinforcing hardening cycle on both sides.
Forecasting U.S.–China AI Rivalry
Analysts, policymakers, and strategists evaluating U.S.–China dynamics in AI should also take the following into account.
Early-Warning Indicator for Target Selection:
Highlighting how policy discourse surrounding kinetic and offensive cyber options against compute infrastructure has moved into mainstream military war-gaming provides an early warning that data centers, subsea cables, and fab energy grids are transitioning from commercial infrastructure to primary targets in military planning.
Exposing the "Containment Trap":
Aggressive export controls (BIS chokepoints) have so far failed to freeze Chinese capability; instead, they have forced China into rapid full-stack indigenization (Huawei Ascend, SMIC) and asymmetric workarounds (algorithmic parameter efficiency, embodied industrial robotics).
De-escalation Diagnostic:
A realistic metric for checking whether U.S.–China AI relations are improving or deteriorating should be devised and accorded priority. If policymakers continue to conflate commercial frontier LLMs/AGI development with existential counter-force threats, any formal bilateral agreements on "AI nuclear C2" or "joint safety guidelines" will remain unachievable.
Conclusion
What is visible in western media and policy circles is the significant disconnect between internal defense-policy planning and public media discourse.
While Western private sector and media driven commentary worries about "accidental" AI war triggered by algorithmic glitches, it routinely overlooks the risk of deliberate, conventional preemption—a military decision driven by the fear that an adversary is months away from an unmatchable AGI breakthrough. If AGI is treated as an existential threat capable of permanently deciding the global balance of power, the strategic logic of counter-force preemption mirrors Cold War nuclear doctrine but without the safeguards necessary to ensure the Doomsday clock does not strike 12.
本文观点，不代表《东方日报》立场。
要看最快最熱資訊，請來Follow我們 《東方日報》WhatsApp Channel.
追踪东方日报社交平台
Telegram Xiaohongshu logo - Brandlogos.net追看热门新闻资讯，请下载东方日报APP
Google Play：https://odn.my/android
HUAWEI AppGallery：https://odn.my/appgallery
评论 (0)